This document proposes an interactive classification system called InClas that involves human experts to improve automatic classification results. It summarizes that traditional automatic classification methods are not always appropriate, especially in domains with small training data, complex attributes, or when experts do not fully trust automated outputs. InClas detects uncertain and unclassified objects for human experts to review and provides a transparent model that the expert can update. Experimental results show InClas can significantly reduce misclassified objects compared to fully automatic classifiers. The document recommends InClas for domains where human experts are available and attributes are understandable, or when obtaining correct classifications for many objects is important and expert time investment is acceptable.